Combinatorial testing is a practical software approach that has been widely used in practice. Most research and applications of such focus on N-way combinatorial provides minimum coverage all interactions among factor. However, the strengths different may not be fixed integer N, but variable. Therefore, variable strength necessary applications. Existing testing, which allows some have higher than others, limitation higher-strength must disjoint. To avoid limitation, an improved approach,...

Combinatorial testing could detect the faults triggered by interactions among factors in software. But many cases, pair-wise, N-way and even variable strength combinatorial may lead test suite redundancy fault ability decreasing, because these methods do not make sufficient consideration on actual interaction. In this paper, a new interaction relationship based model was proposed to cover factor software extending conventional IO model. The method be more effectively than existed without...

Java exception mechanism can effectively free a program from abnormal exits and help developers locate faults with the tracing stacks. It is necessary to verify whether handling constructs are arranged appropriately. Some approaches have been developed evaluate test sets improve quality of them, so that they raise more number exceptions in programs. Mutation analysis practical method sets. This paper presents some new mutation operators for constructs. Moreover, equivalent mutants be...

We propose a convolutional neural network architecture with k-max pooling layer for semantic modeling of music. The aim music model is to analyze and represent the content purposes classification, discovery, or clustering. used in make it possible pool k most active features, capturing semantic-rich time-varying information about Our takes an input as sequence audio words, where each word associated distributed feature vector that can be fine-tuned by backpropagating errors during training....

In the field of deep learning, people strive to construct high-quality neural networks (DNNs) improve accuracy predicting. As well known, quality training data have great impacts on DNN models, since all models are obtained by using these data. However, there is not any reported systematic study how affects model. To relationships between and model quality, we mainly consider four aspects including Skewed Classes, Sample Complexity, Label Quality, Noisy Data in this paper. We design...

Combinatorial testing has been widely utilized in softwares, e.g. Siemens Suite. This paper aims to investigate the reason why combinatorial works Experiments are designed get MFS(minimal failure-causing schema) for Suite, which used as a benchmark evaluate effectiveness of many techniques. The lowerbound fault-detecting probability T-way test suite each program is calculated by analyzing strengths and number MFS faulty version. Computational results could explain

Combinatorial testing is a specification-based method, which can detect the faults triggered by interaction of factors. For one kind software in interactions only exist between neighbor factors, this paper proposes concept factors combinatorial testing, presents covering array generation algorithms for pair-wise (N=2) coverage, N-way (Nges2) coverage and variable strength proves that arrays generated these three are optimal. Finally we analyze an application scenario, shows approach very practical

Background. Mutation testing is a commonly used defect injection technique for evaluating the effectiveness of test suite. However, it usually computationally expensive. Therefore, many mutation reduction strategies, which aim to reduce number mutants, have been proposed. Problem. It important measure ability strategy maintain suite evaluation. existing evaluation indicators are unable “order-preserving ability”, i.e., what extent score order among suites maintained before and after...

PHP (Hypertext Preprocessor) is a scripting language that has been widely used in web development. This paper conducts an empirical study on bugs PHP. By analyzing 35,921 bug reports, 6524 revisions, and root causes of randomly selected 500 bugs, we find that: (1) Among all the 385 versions involved these there are most 4.0.4, 4.0.6, 4.0.3; Documentation mainly distributed 4.y.z 5.y.z; Security primarily relatively later normal 5.y.z. (2) Documentation, Compile, Scripting Engine packages...

Nowadays, people strive to improve the accuracy of deep learning models. However, very little work has focused on quality data sets. In fact, determines model quality. Therefore, it is important for us make research how affects this paper, we mainly consider four aspects quality, including Dataset Equilibrium, Size, Quality Label, Contamination. We deign experiment MNIST and Cifar-10 try find out influence Experimental results show that all have decisive impact It means decrease in these...

With the widespread of microblog, microblog platform like Sina has become information provider to release mass information. Mining valuable from and making personalized recommendations have a hotspot recent research. But most researches about primarily focus on user's social circle. The recommendation is mainly based friends or users they follow, study how find with similar hobbies unfamiliar circles rare. Therefore, we propose generation model UT-LDA (User Topic Latent Dirichlet Allocation)...

As an effective software testing technique, combinatorial has been gradually applied in various types of test practice. In this case, it is necessary to provide useful tools support the application technique on industrial scenarios, as well academic research for technique. To end, basis results group, a suite developed, whose functions include case generation, optimization, and etc. For requirements from both should be configurable, scalable, modular, This paper gives brief introduction...

Inefficient use of memory may cause bloat, and the bloat lead to performance slowdowns or even crashes. To address problem, lots efforts have been devoted diagnosing especially caused by leaks. However, testing identifying executions potentially containing which act as key steps before diagnosing, are still challenging. In this paper, we introduce a specification tool named MemSpec help programmers test inefficiencies in Java programs. The works oracle automatically determine whether program...

Many test adequacy metrics for deep neural net-works (DNNs) were proposed to measure the quality of testing learning system. The combinatorial coverage neurons was because it considers influence in same layer on next layer. However, is still inconclusive whether beneficial DNNs. We conduct an empirical study MNIST dataset answer how affects fault detection experimental results show a medium or strong correlation between fixed-strength and number adversarial examples Such suggests that...

Background. Software testing is a critical activity for ensuring the quality and reliability of software systems. To evaluate effectiveness different test suites, researchers have developed variety metrics. Problem. However, comparing these metrics challenging due to lack standardized evaluation framework including comprehensive factors. As result, often focus on single factors (e.g., size), which finally leads or even contradictory conclusions. After dozens pieces work in detail, we found...

The combinatorial testing technique can detect software faults caused by interactions of different factors (input parameters), using a small number test cases. After fault was detected, its corresponding minimal failure-causing schemas should be identified to help fix such fault. In this paper, we propose an approach calculate based on the relational tree model. Such could without generating additional Case studies show us that proposed method identify effectively and efficiently.